He thinks your second database is a mistake - and he is building the extension to delete it.
Ask most engineers how to add good search to their app and they will tell you to spin up Elasticsearch next to your Postgres, build a pipeline to keep the two in sync, and pray it holds. Philippe Noël's answer is shorter: don't. ParadeDB, the company he co-founded and runs, puts BM25-grade full-text search and analytics inside Postgres itself, as an extension named pg_search. No second cluster. No ETL job that breaks at 3 a.m.
The current scene: a tiny office, four people, and a piece of open-source software quietly burrowing into the databases of Alibaba, Modern Treasury, Bilt Rewards, and TCDI. ParadeDB is the kind of company that sounds modest until you notice who is running it on. In July 2025 Craft Ventures led a $12 million Series A, with Y Combinator following on, pushing total funding to roughly $14 million. The plan for the money is unglamorous on purpose: hire from four people to ten-plus, sharpen the UI, deepen the analytics. Noël is not in a hurry to look big.
That restraint is a position, not an accident. "We prefer to pay more for fewer, highly skilled people than less for more, less-skilled people," he has said. His earlier team spanned the US, Brazil, the UK, and India, hired on one rule - "We just wanted to get the best people, wherever they live." The hard part, he admits, was never the talent. It was knitting a culture across that many time zones.
The idea for ParadeDB did not arrive in a flash. It arrived as an annoyance. While running his previous startup, Noël kept hitting the same wall: Postgres held all the data, but searching it well meant bolting on a separate engine. "The two databases are not meant to work together," he says. "That breaks all the time." When that startup wound down and he drifted into contracting, the annoyance was still there, unsolved, waiting. So he and co-founder Ming Ying started with full-text search, then expanded into real-time analytics, leaning on open-source building blocks - inverted indexes, Postgres internals, a hybrid of row and column storage - rather than reinventing the wheel.
Before all of this there was Whist. Founded in 2019, it was a cloud-hybrid browser with an oddly specific origin: letting animators run heavy creative software like Adobe remotely, from any device. It raised about $3 million and earned a following. Then the ground shifted underneath it. Personal devices got faster, security tooling got better, and the case for a remote browser thinned out. Whist closed at the end of 2022. Noël does not narrate this as a tragedy. He treats quitting as a skill - knowing when a thesis has expired is part of the job, not a failure of nerve.
The roots run further back, into the woods of Quebec. He grew up in Rivière-du-Loup, a small town on the south shore of the Saint Lawrence, before heading south for university. At Harvard he studied computer science and economics, graduating in 2020. Along the way he interned at Premier Tech, automating quality control for industrial robotics, and at Microsoft, improving file server management inside Azure's hybrid cloud. The throughline from rural Quebec to enterprise infrastructure is not obvious until you see what he chose to build: plumbing. Unsexy, load-bearing, the kind of thing the whole web quietly leans on.
His go-to-market is the same as his temperament: build in the open, win developers first, let the rest follow. ParadeDB's first paying customer came from his own Y Combinator batch, reeled in through open source and a steady social-media presence rather than a sales team. The philosophy scales to fundraising too. "If you focus on your customers and do well, the VCs will notice and reach out," he says. The advice he gives to YC hopefuls is two words wide and a career deep: "Understand your customers deeply."
What he is betting on is bigger than one extension. Postgres has become the gravitational center of the data world, and Noël's wager is that the era of pairing it with a separate search engine is ending. If he is right, a generation of brittle pipelines disappears, and the second database - the one everyone has quietly resented for years - simply goes away. That is the whole pitch. It is also, to the search-engine business, a small heresy.
pg_search brings text, hybrid, and faceted search to Postgres with BM25 indexing - the ranking algorithm behind serious search engines - so queries run where the data already lives.
It can run as a logical replica of a managed Postgres instance or install straight into a self-hosted one. The pipeline-that-always-breaks simply stops existing.
12+ tokenizers, 20+ languages, boolean queries, top-K results, and aggregates - wired into frameworks like Django, Rails, SQLAlchemy, Drizzle, and EF Core.
Distributed as open source with 1M+ Docker pulls and 200K+ installs, plus integrations on Railway, Render, and DigitalOcean.
Alibaba, Modern Treasury, Bilt Rewards, and TCDI run it in production - enterprise weight on a four-person team.
That Postgres becomes the one database you need, retiring the Postgres-plus-Elasticsearch architecture for good.
Woods kid. He grew up in Rivière-du-Loup, a small town in rural Quebec, before heading to the US for school.
First idea, odd shape. Whist started as a way for animators to run Adobe software remotely from any device.
Customer zero. ParadeDB's first paying customer came from his own YC batch, won through open source and social media.
Quitting as craft. He frames shutting down a startup as a strategic skill, not a defeat.
Soundtrack. He cites Spotify as personally significant - part learning tool, part way to unwind.
Pain to product. The whole company grew out of a search headache he hit while running his last one.